Fitting determinantal point processes to macro base station deployments

Abstract

The macro base station (BS) deployments in modern cellular networks are neither regular nor completely random. We use determinantal point process (DPP) models to study the repulsiveness among macro base stations observed in cellular networks. Three DPP models are fitted to base station location data sets from two major US cities. Hypothesis testing is used to validate the goodness-of-fit for these DPP models. Based on performance metrics including the K-function, the L-function and coverage probability, DPP models are shown to be accurate in modeling real BS deployments. On the contrary, the Poisson point process and perturbed hexagonal grid model are shown to be less realistic. Different DPP models are compared, and several computational properties of these models are also discussed.